Understanding metric-related pitfalls in image analysis validation

Item Type:	Review
Title:	Understanding metric-related pitfalls in image analysis validation
Creators Name:	Reinke, A. and Tizabi, M.D. and Baumgartner, M. and Eisenmann, M. and Heckmann-Nötzel, D. and Kavur, A.E. and Rädsch, T. and Sudre, C.H. and Acion, L. and Antonelli, M. and Arbel, T. and Bakas, S. and Benis, A. and Buettner, F. and Cardoso, M.J. and Cheplygina, V. and Chen, J. and Christodoulou, E. and Cimini, B.A. and Farahani, K. and Ferrer, L. and Galdran, A. and van Ginneken, B. and Glocker, B. and Godau, P. and Hashimoto, D.A. and Hoffman, M.M. and Huisman, M. and Isensee, F. and Jannin, P. and Kahn, C.E. and Kainmueller, D. and Kainz, B. and Karargyris, A. and Kleesiek, J. and Kofler, F. and Kooi, T. and Kopp-Schneider, A. and Kozubek, M. and Kreshuk, A. and Kurc, T. and Landman, B.A. and Litjens, G. and Madani, A. and Maier-Hein, K. and Martel, A.L. and Meijering, E. and Menze, B. and Moons, K.G.M. and Müller, H. and Nichyporuk, B. and Nickel, F. and Petersen, J. and Rafelski, S.M. and Rajpoot, N. and Reyes, M. and Riegler, M.A. and Rieke, N. and Saez-Rodriguez, J. and Sánchez, C.I. and Shetty, S. and Summers, R.M. and Taha, A.A. and Tiulpin, A. and Tsaftaris, S.A. and Van Calster, B. and Varoquaux, G. and Yaniv, Z.R. and Jäger, P.F. and Maier-Hein, L.
Abstract:	Validation metrics are key for tracking scientific progress and bridging the current chasm between artificial intelligence research and its translation into practice. However, increasing evidence shows that, particularly in image analysis, metrics are often chosen inadequately. Although taking into account the individual strengths, weaknesses and limitations of validation metrics is a critical prerequisite to making educated choices, the relevant knowledge is currently scattered and poorly accessible to individual researchers. Based on a multistage Delphi process conducted by a multidisciplinary expert consortium as well as extensive community feedback, the present work provides a reliable and comprehensive common point of access to information on pitfalls related to validation metrics in image analysis. Although focused on biomedical image analysis, the addressed pitfalls generalize across application domains and are categorized according to a newly created, domain-agnostic taxonomy. The work serves to enhance global comprehension of a key topic in image analysis validation.
Keywords:	Cancer, Education, Medical Research, Artificial Intelligence
Source:	Nature Methods
ISSN:	1548-7091
Publisher:	Nature Publishing Group
Volume:	21
Number:	2
Page Range:	182–194
Date:	February 2024
Official Publication:	https://doi.org/10.1038/s41592-023-02150-0
PubMed:	View item in PubMed

Repository Staff Only: item control page